ObjectOSAI Credits and Limits
How the in-product AI is metered on ObjectOS Cloud — one allowance for building and asking, a rolling weekly limit and a five-hour pace, credit packs, the model each plan uses, and what a refusal says.
On ObjectOS Cloud we run the AI model for you, so the in-product AI — the AI Builder and the assistant that answers questions about your data — draws on an AI allowance. On a self-managed ObjectOS Enterprise deployment you bring your own model instead, and none of this page's limits apply: see On ObjectOS Enterprise.
One allowance for all AI work
Building an app, asking about your data and running actions through the assistant all draw on the same allowance. There is no separate budget per kind of work. Each request draws on it in proportion to the work the model does for it, and a more capable model, where your plan offers a choice, draws more.
The allowance is counted per environment. Home in the cloud console shows how much of this week's allowance has been used.
Two rolling windows
Every plan has the same two limits, and both are rolling: usage counts for a fixed length of time and then drops out, so room comes back continuously rather than on a calendar date. Ruled on cloud#2059.
| Window | What it is | Free | Team and above |
|---|---|---|---|
| 7 days | The budget: how much AI work fits in a week | One weekly allowance per environment; a hard limit | A larger allowance per AI seat, multiplied by the organization's AI seats; credit packs raise it |
| 5 hours | The pace: how much fits in a short burst, so one runaway session cannot spend the week | Applies, but sits above Free's weekly allowance | Per AI seat, multiplied by the seats; credit packs do not raise it |
The exact sizes of the allowances are not published while ObjectOS Cloud is in Public Preview. Free's is sized for the whole first experience: building an app, changing it several times, and still asking it questions.
When a limit is reached
The request is refused with HTTP 429, and the chat shows a message that says
which limit it was and when it frees up:
| Situation | Code | Message |
|---|---|---|
| Free, weekly allowance used | AI_DESIGN_QUOTA_EXHAUSTED | You've used this week's free AI quota. It resets on DATE — upgrade to keep going. |
| Paid, weekly allowance used | AI_ALLOWANCE_EXHAUSTED | Your AI allowance for this week is used up. It resets on DATE, or buy a credit pack to continue now. |
| Any plan, five-hour pace reached | AI_DESIGN_QUOTA_EXHAUSTED on Free, AI_ALLOWANCE_EXHAUSTED on a paid plan | You've hit the five-hour AI usage limit. It resets within five hours — then you're good to go. |
The date is when the oldest usage still counted leaves the window; the error's
details.resetsAt carries the same moment as an ISO timestamp. Nothing else in
the environment is affected: your apps, data and integrations keep working.
Buy AI credits
On a paid plan, an owner or admin can top up the weekly allowance with a one-off AI credit pack: open the environment's Settings → Plan and billing, then AI credits → Buy AI Credits. A secure checkout opens in a new tab.
- A credit pack raises the 7-day budget only. The five-hour pace is a rate limit and no purchase lifts it.
- Free does not take credit packs; the way past Free's weekly limit is the upgrade to Team.
- If credit packs are not on sale yet for your organization, the button says so instead of opening a checkout.
Which model answers
Every plan answers with the same default model: paid plans differ from Free by the size of the allowance, not by a stronger model. Where a choice of models is offered to a paid environment, the AI chat lets you pick one, and a more capable model draws the allowance down faster.
AI through your own client is not metered here
An agent you connect over MCP — Claude Code, Cursor and the like — runs on your model and your own tokens. It does not draw on the AI allowance. What bounds it is the environment's MCP request limit (Storage and Limits).
On ObjectOS Enterprise
A self-managed Enterprise deployment brings its own model and key, configured in AI Service, and pays its model provider directly. There are no platform AI credits, no allowance and no metering: the per-AI-seat fee is a license, not a usage bundle. How much AI work your deployment does is limited only by your own provider account.
Next
- Plans and AI Seats — what each plan includes.
- AI Builder — what the AI can build.
Plans and AI Seats
The ObjectOS Cloud plans — Free, Team and Enterprise — the AI seat as the only paid seat, how to upgrade, and where the console shows what your plan leaves out.
Storage and Limits
The usage limits of an ObjectOS Cloud environment — the 1 GB and 20 GB storage allowances, what a full environment stops and never stops, the 80% warning, and the MCP request and draft limits per plan.